7
7
# it under the terms of the GNU General Public License as published by
8
8
# the Free Software Foundation; either version 2 of the License, or
9
9
# (at your option) any later version.
11
11
# This program is distributed in the hope that it will be useful,
12
12
# but WITHOUT ANY WARRANTY; without even the implied warranty of
13
13
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14
14
# GNU General Public License for more details.
16
16
# You should have received a copy of the GNU General Public License
17
17
# along with this program; if not, write to the Free Software
18
18
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20
# Remaing to do is to figure out if get_graph should return a simple
21
# map, or a graph object of some kind.
24
20
"""Versioned text file storage api."""
27
from copy import deepcopy
28
from unittest import TestSuite
31
import bzrlib.errors as errors
22
from bzrlib.lazy_import import lazy_import
23
lazy_import(globals(), """
33
from bzrlib.transport.memory import MemoryTransport
36
from cStringIO import StringIO
32
38
from bzrlib.inter import InterObject
33
from bzrlib.symbol_versioning import *
34
from bzrlib.transport.memory import MemoryTransport
35
from bzrlib.tsort import topo_sort
39
from bzrlib.textmerge import TextMerge
39
42
class VersionedFile(object):
78
77
"""Returns whether version is present."""
79
78
raise NotImplementedError(self.has_version)
81
def add_delta(self, version_id, parents, delta_parent, sha1, noeol, delta):
82
"""Add a text to the versioned file via a pregenerated delta.
84
:param version_id: The version id being added.
85
:param parents: The parents of the version_id.
86
:param delta_parent: The parent this delta was created against.
87
:param sha1: The sha1 of the full text.
88
:param delta: The delta instructions. See get_delta for details.
90
self._check_write_ok()
91
if self.has_version(version_id):
92
raise errors.RevisionAlreadyPresent(version_id, self)
93
return self._add_delta(version_id, parents, delta_parent, sha1, noeol, delta)
95
def _add_delta(self, version_id, parents, delta_parent, sha1, noeol, delta):
96
"""Class specific routine to add a delta.
98
This generic version simply applies the delta to the delta_parent and
101
# strip annotation from delta
103
for start, stop, delta_len, delta_lines in delta:
104
new_delta.append((start, stop, delta_len, [text for origin, text in delta_lines]))
105
if delta_parent is not None:
106
parent_full = self.get_lines(delta_parent)
109
new_full = self._apply_delta(parent_full, new_delta)
110
# its impossible to have noeol on an empty file
111
if noeol and new_full[-1][-1] == '\n':
112
new_full[-1] = new_full[-1][:-1]
113
self.add_lines(version_id, parents, new_full)
115
def add_lines(self, version_id, parents, lines, parent_texts=None):
80
def add_lines(self, version_id, parents, lines, parent_texts=None,
81
left_matching_blocks=None, nostore_sha=None, random_id=False,
116
83
"""Add a single text on top of the versioned file.
118
85
Must raise RevisionAlreadyPresent if the new version is
121
88
Must raise RevisionNotPresent if any of the given parents are
122
89
not present in file history.
91
:param lines: A list of lines. Each line must be a bytestring. And all
92
of them except the last must be terminated with \n and contain no
93
other \n's. The last line may either contain no \n's or a single
94
terminated \n. If the lines list does meet this constraint the add
95
routine may error or may succeed - but you will be unable to read
96
the data back accurately. (Checking the lines have been split
97
correctly is expensive and extremely unlikely to catch bugs so it
98
is not done at runtime unless check_content is True.)
123
99
:param parent_texts: An optional dictionary containing the opaque
124
representations of some or all of the parents of
125
version_id to allow delta optimisations.
126
VERY IMPORTANT: the texts must be those returned
127
by add_lines or data corruption can be caused.
128
:return: An opaque representation of the inserted version which can be
129
provided back to future add_lines calls in the parent_texts
100
representations of some or all of the parents of version_id to
101
allow delta optimisations. VERY IMPORTANT: the texts must be those
102
returned by add_lines or data corruption can be caused.
103
:param left_matching_blocks: a hint about which areas are common
104
between the text and its left-hand-parent. The format is
105
the SequenceMatcher.get_matching_blocks format.
106
:param nostore_sha: Raise ExistingContent and do not add the lines to
107
the versioned file if the digest of the lines matches this.
108
:param random_id: If True a random id has been selected rather than
109
an id determined by some deterministic process such as a converter
110
from a foreign VCS. When True the backend may choose not to check
111
for uniqueness of the resulting key within the versioned file, so
112
this should only be done when the result is expected to be unique
114
:param check_content: If True, the lines supplied are verified to be
115
bytestrings that are correctly formed lines.
116
:return: The text sha1, the number of bytes in the text, and an opaque
117
representation of the inserted version which can be provided
118
back to future add_lines calls in the parent_texts dictionary.
132
120
self._check_write_ok()
133
return self._add_lines(version_id, parents, lines, parent_texts)
121
return self._add_lines(version_id, parents, lines, parent_texts,
122
left_matching_blocks, nostore_sha, random_id, check_content)
135
def _add_lines(self, version_id, parents, lines, parent_texts):
124
def _add_lines(self, version_id, parents, lines, parent_texts,
125
left_matching_blocks, nostore_sha, random_id, check_content):
136
126
"""Helper to do the class specific add_lines."""
137
127
raise NotImplementedError(self.add_lines)
139
129
def add_lines_with_ghosts(self, version_id, parents, lines,
130
parent_texts=None, nostore_sha=None, random_id=False,
141
132
"""Add lines to the versioned file, allowing ghosts to be present.
143
This takes the same parameters as add_lines.
134
This takes the same parameters as add_lines and returns the same.
145
136
self._check_write_ok()
146
137
return self._add_lines_with_ghosts(version_id, parents, lines,
138
parent_texts, nostore_sha, random_id, check_content)
149
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts):
140
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
141
nostore_sha, random_id, check_content):
150
142
"""Helper to do class specific add_lines_with_ghosts."""
151
143
raise NotImplementedError(self.add_lines_with_ghosts)
189
204
raise NotImplementedError(self.create_empty)
191
def fix_parents(self, version, new_parents):
192
"""Fix the parents list for version.
194
This is done by appending a new version to the index
195
with identical data except for the parents list.
196
the parents list must be a superset of the current
199
self._check_write_ok()
200
return self._fix_parents(version, new_parents)
202
def _fix_parents(self, version, new_parents):
203
"""Helper for fix_parents."""
204
raise NotImplementedError(self.fix_parents)
206
def get_delta(self, version):
207
"""Get a delta for constructing version from some other version.
209
:return: (delta_parent, sha1, noeol, delta)
210
Where delta_parent is a version id or None to indicate no parent.
212
raise NotImplementedError(self.get_delta)
214
def get_deltas(self, versions):
215
"""Get multiple deltas at once for constructing versions.
217
:return: dict(version_id:(delta_parent, sha1, noeol, delta))
218
Where delta_parent is a version id or None to indicate no parent, and
219
version_id is the version_id created by that delta.
222
for version in versions:
223
result[version] = self.get_delta(version)
206
def get_format_signature(self):
207
"""Get a text description of the data encoding in this file.
211
raise NotImplementedError(self.get_format_signature)
213
def make_mpdiffs(self, version_ids):
214
"""Create multiparent diffs for specified versions."""
215
knit_versions = set()
216
for version_id in version_ids:
217
knit_versions.add(version_id)
218
knit_versions.update(self.get_parents(version_id))
219
lines = dict(zip(knit_versions,
220
self._get_lf_split_line_list(knit_versions)))
222
for version_id in version_ids:
223
target = lines[version_id]
224
parents = [lines[p] for p in self.get_parents(version_id)]
226
left_parent_blocks = self._extract_blocks(version_id,
229
left_parent_blocks = None
230
diffs.append(multiparent.MultiParent.from_lines(target, parents,
234
def _extract_blocks(self, version_id, source, target):
237
def add_mpdiffs(self, records):
238
"""Add mpdiffs to this VersionedFile.
240
Records should be iterables of version, parents, expected_sha1,
241
mpdiff. mpdiff should be a MultiParent instance.
243
# Does this need to call self._check_write_ok()? (IanC 20070919)
245
mpvf = multiparent.MultiMemoryVersionedFile()
247
for version, parent_ids, expected_sha1, mpdiff in records:
248
versions.append(version)
249
mpvf.add_diff(mpdiff, version, parent_ids)
250
needed_parents = set()
251
for version, parent_ids, expected_sha1, mpdiff in records:
252
needed_parents.update(p for p in parent_ids
253
if not mpvf.has_version(p))
254
for parent_id, lines in zip(needed_parents,
255
self._get_lf_split_line_list(needed_parents)):
256
mpvf.add_version(lines, parent_id, [])
257
for (version, parent_ids, expected_sha1, mpdiff), lines in\
258
zip(records, mpvf.get_line_list(versions)):
259
if len(parent_ids) == 1:
260
left_matching_blocks = list(mpdiff.get_matching_blocks(0,
261
mpvf.get_diff(parent_ids[0]).num_lines()))
263
left_matching_blocks = None
264
_, _, version_text = self.add_lines(version, parent_ids, lines,
265
vf_parents, left_matching_blocks=left_matching_blocks)
266
vf_parents[version] = version_text
267
for (version, parent_ids, expected_sha1, mpdiff), sha1 in\
268
zip(records, self.get_sha1s(versions)):
269
if expected_sha1 != sha1:
270
raise errors.VersionedFileInvalidChecksum(version)
272
def get_sha1(self, version_id):
273
"""Get the stored sha1 sum for the given revision.
275
:param version_id: The name of the version to lookup
277
raise NotImplementedError(self.get_sha1)
279
def get_sha1s(self, version_ids):
280
"""Get the stored sha1 sums for the given revisions.
282
:param version_ids: The names of the versions to lookup
283
:return: a list of sha1s in order according to the version_ids
285
raise NotImplementedError(self.get_sha1s)
226
287
def get_suffixes(self):
227
288
"""Return the file suffixes associated with this versioned file."""
353
def iter_lines_added_or_present_in_versions(self, version_ids=None):
423
def iter_lines_added_or_present_in_versions(self, version_ids=None,
354
425
"""Iterate over the lines in the versioned file from version_ids.
356
This may return lines from other versions, and does not return the
357
specific version marker at this point. The api may be changed
358
during development to include the version that the versioned file
359
thinks is relevant, but given that such hints are just guesses,
360
its better not to have it if we dont need it.
427
This may return lines from other versions. Each item the returned
428
iterator yields is a tuple of a line and a text version that that line
429
is present in (not introduced in).
431
Ordering of results is in whatever order is most suitable for the
432
underlying storage format.
434
If a progress bar is supplied, it may be used to indicate progress.
435
The caller is responsible for cleaning up progress bars (because this
362
438
NOTES: Lines are normalised: they will all have \n terminators.
363
439
Lines are returned in arbitrary order.
441
:return: An iterator over (line, version_id).
365
443
raise NotImplementedError(self.iter_lines_added_or_present_in_versions)
445
def iter_parents(self, version_ids):
446
"""Iterate through the parents for many version ids.
448
:param version_ids: An iterable yielding version_ids.
449
:return: An iterator that yields (version_id, parents). Requested
450
version_ids not present in the versioned file are simply skipped.
451
The order is undefined, allowing for different optimisations in
452
the underlying implementation.
454
for version_id in version_ids:
456
yield version_id, tuple(self.get_parents(version_id))
457
except errors.RevisionNotPresent:
367
460
def transaction_finished(self):
368
461
"""The transaction that this file was opened in has finished.
402
474
Weave lines present in none of them are skipped entirely.
404
inc_a = set(self.get_ancestry([ver_a]))
405
inc_b = set(self.get_ancestry([ver_b]))
406
inc_c = inc_a & inc_b
408
for lineno, insert, deleteset, line in self.walk([ver_a, ver_b]):
409
if deleteset & inc_c:
410
# killed in parent; can't be in either a or b
411
# not relevant to our work
412
yield 'killed-base', line
413
elif insert in inc_c:
414
# was inserted in base
415
killed_a = bool(deleteset & inc_a)
416
killed_b = bool(deleteset & inc_b)
417
if killed_a and killed_b:
418
yield 'killed-both', line
420
yield 'killed-a', line
422
yield 'killed-b', line
424
yield 'unchanged', line
425
elif insert in inc_a:
426
if deleteset & inc_a:
427
yield 'ghost-a', line
431
elif insert in inc_b:
432
if deleteset & inc_b:
433
yield 'ghost-b', line
477
killed-base Dead in base revision
478
killed-both Killed in each revision
481
unchanged Alive in both a and b (possibly created in both)
484
ghost-a Killed in a, unborn in b
485
ghost-b Killed in b, unborn in a
486
irrelevant Not in either revision
488
raise NotImplementedError(VersionedFile.plan_merge)
490
def weave_merge(self, plan, a_marker=TextMerge.A_MARKER,
491
b_marker=TextMerge.B_MARKER):
492
return PlanWeaveMerge(plan, a_marker, b_marker).merge_lines()[0]
495
class _PlanMergeVersionedFile(object):
496
"""A VersionedFile for uncommitted and committed texts.
498
It is intended to allow merges to be planned with working tree texts.
499
It implements only the small part of the VersionedFile interface used by
500
PlanMerge. It falls back to multiple versionedfiles for data not stored in
501
_PlanMergeVersionedFile itself.
504
def __init__(self, file_id, fallback_versionedfiles=None):
507
:param file_id: Used when raising exceptions.
508
:param fallback_versionedfiles: If supplied, the set of fallbacks to
509
use. Otherwise, _PlanMergeVersionedFile.fallback_versionedfiles
510
can be appended to later.
512
self._file_id = file_id
513
if fallback_versionedfiles is None:
514
self.fallback_versionedfiles = []
516
self.fallback_versionedfiles = fallback_versionedfiles
520
def plan_merge(self, ver_a, ver_b, base=None):
521
"""See VersionedFile.plan_merge"""
522
from bzrlib.merge import _PlanMerge
524
return _PlanMerge(ver_a, ver_b, self).plan_merge()
525
old_plan = list(_PlanMerge(ver_a, base, self).plan_merge())
526
new_plan = list(_PlanMerge(ver_a, ver_b, self).plan_merge())
527
return _PlanMerge._subtract_plans(old_plan, new_plan)
529
def plan_lca_merge(self, ver_a, ver_b, base=None):
530
from bzrlib.merge import _PlanLCAMerge
531
graph = self._get_graph()
532
new_plan = _PlanLCAMerge(ver_a, ver_b, self, graph).plan_merge()
535
old_plan = _PlanLCAMerge(ver_a, base, self, graph).plan_merge()
536
return _PlanLCAMerge._subtract_plans(list(old_plan), list(new_plan))
538
def add_lines(self, version_id, parents, lines):
539
"""See VersionedFile.add_lines
541
Lines are added locally, not fallback versionedfiles. Also, ghosts are
542
permitted. Only reserved ids are permitted.
544
if not revision.is_reserved_id(version_id):
545
raise ValueError('Only reserved ids may be used')
547
raise ValueError('Parents may not be None')
549
raise ValueError('Lines may not be None')
550
self._parents[version_id] = parents
551
self._lines[version_id] = lines
553
def get_lines(self, version_id):
554
"""See VersionedFile.get_ancestry"""
555
lines = self._lines.get(version_id)
556
if lines is not None:
558
for versionedfile in self.fallback_versionedfiles:
560
return versionedfile.get_lines(version_id)
561
except errors.RevisionNotPresent:
564
raise errors.RevisionNotPresent(version_id, self._file_id)
566
def get_ancestry(self, version_id, topo_sorted=False):
567
"""See VersionedFile.get_ancestry.
569
Note that this implementation assumes that if a VersionedFile can
570
answer get_ancestry at all, it can give an authoritative answer. In
571
fact, ghosts can invalidate this assumption. But it's good enough
572
99% of the time, and far cheaper/simpler.
574
Also note that the results of this version are never topologically
575
sorted, and are a set.
578
raise ValueError('This implementation does not provide sorting')
579
parents = self._parents.get(version_id)
581
for vf in self.fallback_versionedfiles:
583
return vf.get_ancestry(version_id, topo_sorted=False)
584
except errors.RevisionNotPresent:
437
# not in either revision
438
yield 'irrelevant', line
440
yield 'unchanged', '' # terminator
442
def weave_merge(self, plan, a_marker='<<<<<<< \n', b_marker='>>>>>>> \n'):
587
raise errors.RevisionNotPresent(version_id, self._file_id)
588
ancestry = set([version_id])
589
for parent in parents:
590
ancestry.update(self.get_ancestry(parent, topo_sorted=False))
593
def get_parents(self, version_id):
594
"""See VersionedFile.get_parents"""
595
parents = self._parents.get(version_id)
596
if parents is not None:
598
for versionedfile in self.fallback_versionedfiles:
600
return versionedfile.get_parents(version_id)
601
except errors.RevisionNotPresent:
604
raise errors.RevisionNotPresent(version_id, self._file_id)
606
def _get_graph(self):
607
from bzrlib.graph import (
610
_StackedParentsProvider,
612
from bzrlib.repofmt.knitrepo import _KnitParentsProvider
613
parent_providers = [DictParentsProvider(self._parents)]
614
for vf in self.fallback_versionedfiles:
615
parent_providers.append(_KnitParentsProvider(vf))
616
return Graph(_StackedParentsProvider(parent_providers))
619
class PlanWeaveMerge(TextMerge):
620
"""Weave merge that takes a plan as its input.
622
This exists so that VersionedFile.plan_merge is implementable.
623
Most callers will want to use WeaveMerge instead.
626
def __init__(self, plan, a_marker=TextMerge.A_MARKER,
627
b_marker=TextMerge.B_MARKER):
628
TextMerge.__init__(self, a_marker, b_marker)
631
def _merge_struct(self):
445
634
ch_a = ch_b = False
446
# TODO: Return a structured form of the conflicts (e.g. 2-tuples for
447
# conflicted regions), rather than just inserting the markers.
449
# TODO: Show some version information (e.g. author, date) on
450
# conflicted regions.
636
def outstanding_struct():
637
if not lines_a and not lines_b:
639
elif ch_a and not ch_b:
642
elif ch_b and not ch_a:
644
elif lines_a == lines_b:
647
yield (lines_a, lines_b)
452
649
# We previously considered either 'unchanged' or 'killed-both' lines
453
650
# to be possible places to resynchronize. However, assuming agreement
454
# on killed-both lines may be too agressive. -- mbp 20060324
455
for state, line in plan:
651
# on killed-both lines may be too aggressive. -- mbp 20060324
652
for state, line in self.plan:
456
653
if state == 'unchanged':
457
654
# resync and flush queued conflicts changes if any
458
if not lines_a and not lines_b:
460
elif ch_a and not ch_b:
462
for l in lines_a: yield l
463
elif ch_b and not ch_a:
464
for l in lines_b: yield l
465
elif lines_a == lines_b:
466
for l in lines_a: yield l
469
for l in lines_a: yield l
471
for l in lines_b: yield l
655
for struct in outstanding_struct():
476
659
ch_a = ch_b = False
478
661
if state == 'unchanged':
481
664
elif state == 'killed-a':
483
666
lines_b.append(line)
582
class InterVersionedFileTestProviderAdapter(object):
583
"""A tool to generate a suite testing multiple inter versioned-file classes.
585
This is done by copying the test once for each interversionedfile provider
586
and injecting the transport_server, transport_readonly_server,
587
versionedfile_factory and versionedfile_factory_to classes into each copy.
588
Each copy is also given a new id() to make it easy to identify.
591
def __init__(self, transport_server, transport_readonly_server, formats):
592
self._transport_server = transport_server
593
self._transport_readonly_server = transport_readonly_server
594
self._formats = formats
596
def adapt(self, test):
598
for (interversionedfile_class,
599
versionedfile_factory,
600
versionedfile_factory_to) in self._formats:
601
new_test = deepcopy(test)
602
new_test.transport_server = self._transport_server
603
new_test.transport_readonly_server = self._transport_readonly_server
604
new_test.interversionedfile_class = interversionedfile_class
605
new_test.versionedfile_factory = versionedfile_factory
606
new_test.versionedfile_factory_to = versionedfile_factory_to
607
def make_new_test_id():
608
new_id = "%s(%s)" % (new_test.id(), interversionedfile_class.__name__)
609
return lambda: new_id
610
new_test.id = make_new_test_id()
611
result.addTest(new_test)
615
def default_test_list():
616
"""Generate the default list of interversionedfile permutations to test."""
617
from bzrlib.weave import WeaveFile
618
from bzrlib.knit import KnitVersionedFile
620
# test the fallback InterVersionedFile from weave to annotated knits
621
result.append((InterVersionedFile,
624
for optimiser in InterVersionedFile._optimisers:
625
result.append((optimiser,
626
optimiser._matching_file_factory,
627
optimiser._matching_file_factory
629
# if there are specific combinations we want to use, we can add them
775
def _get_source_version_ids(self, version_ids, ignore_missing):
776
"""Determine the version ids to be used from self.source.
778
:param version_ids: The caller-supplied version ids to check. (None
779
for all). If None is in version_ids, it is stripped.
780
:param ignore_missing: if True, remove missing ids from the version
781
list. If False, raise RevisionNotPresent on
782
a missing version id.
783
:return: A set of version ids.
785
if version_ids is None:
786
# None cannot be in source.versions
787
return set(self.source.versions())
790
return set(self.source.versions()).intersection(set(version_ids))
792
new_version_ids = set()
793
for version in version_ids:
796
if not self.source.has_version(version):
797
raise errors.RevisionNotPresent(version, str(self.source))
799
new_version_ids.add(version)
800
return new_version_ids
added
removed
bzrlib/versionedfile.py
